MIT-CTP/5776

The complex Liouville string

Scott Collier
Center for Theoretical Physics, Massachusetts Institute of Technology, Cambridge, MA 02139, USA

Lorenz Eberhardt
Institute for Theoretical Physics, University of Amsterdam, 1098 XH Amsterdam, Netherlands

Beatrix Mühlmann
School of Natural Sciences, Institute for Advanced Study, Princeton, NJ 08540, USA and
arXiv:2409.17246v1 [hep-th] 25 Sep 2024

Department of Physics, McGill University, Montréal, QC, H3A 2T8, Canada

Victor A. Rodriguez
Department of Physics, University of California, Santa Barbara, CA 93106, USA and
Joseph Henry Laboratories, Princeton University, Princeton, NJ 08544, USA
(Dated: September 27, 2024)
We introduce the complex Liouville string, a solvable string theory defined by coupling two Liouville
theories with complex conjugate central charges c ∈ 13 + iR on the worldsheet. We compute its
amplitudes from first principles and establish a duality with a double-scaled two-matrix integral.
We also analyze general worldsheet boundaries and non-perturbative effects in the genus expansion.
This theory is capable of capturing aspects of de Sitter quantum gravity in both two and three
dimensions, which leads to a precise version of de Sitter holography.

INTRODUCTION actly from first principles by employing analytic boot-

strap techniques. It admits a dual description in terms
Solvable yet interesting models of quantum gravity are of a random two-matrix model, and moreover describes a
rare, but yield valuable insights. A common paradigm for theory of 2d quantum gravity with vacua of both positive
two-dimensional theories of gravity is to define them as a and negative cosmological constants. Thus it provides
critical worldsheet string theory, allowing for the applica- an exact version of holography involving de Sitter (dS)
tion of extensive technical machinery. Notable examples spacetimes, which is reflected in several unusual proper-
include the (p, q)-minimal string [1–4] or the Virasoro ties of the worldsheet theory and of the matrix model.
minimal string [5], both of which admit a dual description The theory allows us to consider observables with gen-
in terms of double-scaled matrix integrals—providing a eral boundaries from the worldsheet, and to study (dou-
version of holographic duality for these theories. The bly) non-perturbative effects in the sum over topologies
presentation as a worldsheet string theory, or as a ma- in detail. Finally, the string amplitudes of (1) can also
trix integral, also provides a non-perturbative definition be reinterpreted as cosmological correlators of 3d pure
of the corresponding 2d model of quantum gravity. de Sitter gravity with non-dynamical massive particles.
These string theories can also be viewed as more tradi- This interpretation gives a precise version of a de Sitter/
tional theories of gravity, consisting of a metric coupled matrix model correspondence. In this letter, we give an
to a dilaton scalar field. In this formulation, the holo- overview of these aspects of CLS and refer to the papers
graphic duality was reinterpreted as a duality between [7–10] for further details.
‘pure’ quantum gravity and a random matrix model [6].
In this letter, we announce a new such model which is
WORLDSHEET THEORY
far richer than previous ones, but remains solvable. On
the worldsheet, it is defined by coupling two Liouville
theories with complex central charges, Definition. The worldsheet theory is defined by cou-
pling two complex conjugate Liouville CFTs. We employ
Liouville CFT (Liouville CFT)∗ bc-ghosts the standard parameterization c = 1 + 6(b + b−1 )2 of the
+ ⊕ ⊕ . (1)
c ∈ 13 + iR+ c− ∈ 13 − iR+ c = −26 central charge [11–14]. c± ∈ 13 + iR in (1) corresponds
πi
to b− = −ib+ ∈ e− 4 R. The combined worldsheet the-
We refer to this theory as the complex Liouville string ory can be endowed with a non-unitary inner product
(CLS). This theory is very interesting from a variety of via the reality condition (L+ † − ±
n ) = L−n where Ln are the
perspectives and holds a surprising number of lessons Virasoro generators of the two Liouville theories. Vertex
in stock. The string amplitudes can be computed ex- operators in the full worldsheet theory can be written as
 2

a product of the Liouville vertex operators Vp++ Vp−− , with vertex operators, which in our normalization read
conformal weights given by h± = (c± − 1)/24 − (p±)2 for −1
(b )
each primary. The reality condition implies h− = (h+ )∗ , Ag,n (p) = (−1)n A(b)
g,n (p) (duality) , (5a)
while the mass-shell condition imposes h+ + h− = 1. A(−ib) n (b)
πi g,n (ip) = (−i) Ag,n (p) (swap) , (5b)
This is equivalent to p− = ip+ , p+ ∈ e− 4 R+ . On-shell
vertex operators of CLS including the ghosts are thus A(b)
g,n (−p1 , p2 , . . . , pn ) = −A(b)
g,n (p) (reflection) , (5c)
Vp ≡ Nb (p)cc̃Vp+ Vip− , where Nb (p) reflects an arbitrary
as well as permutation symmetry in the Liouville mo-
normalization. We can then define perturbative string
menta. The perhaps unusual minus sign in (5c) follows
amplitudes as
from a convenient normalization.
Z  3g−3+n
Y n
Y  Dilaton equation and higher equations of motion.
(b)
A(b)
g,n (p) ≡ CΣg Bk Bek Vpj , (2) Further constraints on the string amplitudes can be de-
Mg,n k=1 j=1 Σg,n rived when specializing one of the momenta to a degen-
erate value pj = 21 (mb − nb−1 ), m, n ∈ Z. One can use
where bold letters refer to tuples p = (p1 , . . . , pn ). Mg,n the technology of higher equations of motion developed
denotes the moduli space of a genus g Riemann surface in the context of minimal string theory [16, 17] to write
with n punctures and Bk denotes the 3g−3+n zero modes the integrand in (2) as a total derivative and localize it
of the b-ghost. The correlation functions in (2) can be to a sublocus in Mg,n . The simplest such relation is the
obtained from a conformal block expansion of Liouville dilaton equation,
theory, which is solved via the DOZZ formula for the
 n q
X 
structure constants [12, 13]. Contrary to the standard (b) Qχg,n
bosonic string in 26 dimensions, the moduli space integral lim Ag,n+1 (p, p′ ) = 2 + p2j A(b)
g,n (p) , (6)
b
p′ → 12 Q j=1
in (2) is absolutely convergent in CLS.
The full string amplitudes are obtained by summing b = b−1 − b and χg,n = 2 − 2g − n.
with Q = b−1 + b, Q
over genera with a string coupling constant gs = e−S0
The square root appears on the right hand side since
X the string amplitude is analytically continued outside its
A(b)
n (S0 ; p) = eS0 (2−2g−n) A(b)
g,n (p) . (3)
initial domain of definition where it develops branch cuts
g≥0
as discussed above.
(b)
Analyticity. The amplitudes (2) are intially defined Analytic bootstrap. Except for A0,3 (p1 , p2 , p3 ) which
πi πi
for pj ∈ e− 4 R+ and b ∈ e 4 R+ , but may be analytically does not involve moduli integration, it is challenging to
continued. On a certain open neighborhood of this real evaluate (2) from first principles since both Mg,n as well
locus, the integral over Mg,n is still absolutely convergent as the Virasoro conformal blocks that are necessary to
(b) define the worldsheet correlators are very complicated.
and thus Ag,n (p) is an analytic function there. Eventu- (b)
ally, however, the Liouville correlators develop resonance Nonetheless, Ag,n can be inferred from the various prop-
poles when analytically continued outside the physical erties mentioned above. This constitutes an analytic
region [13, 15]; this happens for bootstrap problem. For (g, n) = (0, 4) and (g, n) = (1, 1),
it has a unique solution, provided one assumes subexpo-
n (b)
X n nential growth on Ag,n (p) for large values of |pj |, which
±pj = rb + sb−1 , |r|, |s| ∈ Z≥0 + g − 1 + (4) we were not able to establish analytically. We will present
j=1
2
the result below, see eq. (13). For larger values of (g, n),
for any choice of signs. These poles carry over to the we have fewer constraints and the bootstrap problem be-
(b) comes more complex. Therefore we did not try to solve
string amplitudes Ag,n . Furthermore, the Liouville cor-
it exhaustively, but we will check below that the gen-
relators can pick up discrete terms from poles crossing
eral conjecture (13) is consistent with all the constraints
the OPE contour of the conformal block expansion un-
that we derived from the worldsheet. For (g, n) = (0, 4)
der analytic continuation. These spoil convergence over
(b) and (1, 1) we can evaluate the string amplitudes by direct
Mg,n ; and while Ag,n (p) can be analytically continued numerical integration over moduli space, finding results
(b)
further, the string amplitude Ag,n (p) develops branch consistent with the analytic formula within < 10−4 %.
cuts. The discontinuity through the cut is computed
from a local analysis near the corresponding degenera-
tion of the worldsheet and can be expressed in terms of MATRIX MODEL
string amplitudes with lower values of 3g − 3 + n.
Symmetries. The definition (2) has a number of ob- The outputs of the analytic bootstrap are remarkably
vious symmetries stemming from the ambiguity of the compact formulas for the string amplitudes whose under-
parameterization of the central charge through b, inter- lying simplicity is obscured by their origin as worldsheet
changing of the theories, and reflection symmetry of the moduli space integrals (2). This suggests the existence
of a dual matrix model description.
 3

Two-matrix model and topological recursion. Double- From the spectral curve (10) one may extract the lead-
scaled two-matrix integrals are familiar in the physics lit- ing density of eigenvalues of the first matrix by comput-
erature as the conjectured dual descriptions of the (p, q) ing the discontinuity of R0,1 (x):
minimal string [18–21]. The dual of the CLS is conceptu-
ally similar to this family of dualities but with important 2

ρ0 (x) = sinh(−πib2 ) sin −ib2 cosh−1 ( x2 ) . (11)
technical differences. The relevant class of two-matrix π
integrals is defined by integrating a pair of N × N Her- This eigenvalue density is manifestly positive in a neigh-
mitian matrices weighted by suitable potentials together borhood of the edge x = 2, however it oscillates on suffi-
with a minimal coupling ciently large scales, rendering the non-perturbative defi-
Z nition subtle as discussed below.
h · i = [dM1 ][dM2 ] ( · )e−N tr(V1 (M1 )+V2 (M2 )−M1 M2 ) . (7) General string amplitudes. The dictionary between
(b)
the differentials ωg,n and the string amplitudes Ag,n takes
The principal observables in such matrix models are re- the form
solvents, R(x) ≡ tr(x − M1 )−1 , which exhibit a pole Z ∞Y n √
when x coincides with an eigenvalue of M1 . The re- 2πipj dpj exp(2πipj zj )dzj (b)
ωg,n (z) = √ Ag,n (p) ,
lation between the matrix model and CLS arises in a 0 j=1
zj
double-scaling limit, where we take N → ∞ and zoom (12)
in on a particular region of the eigenvalue distribution of
the matrices. Connected products of resolvents admit a which can be inverted via an inverse Laplace transform.
topological expansion One can use the relation between the quantities ωg,n
and intersection numbers on compactified moduli space
DY
n E ∞
X (b)
R(xj ) = Rg,n (x1 , . . . , xn ) eS0 (2−2g−n) , (8) Mg,n to derive a general expression for Ag,n ,
c
j=1 g=0
X Z ′Y
1
A(b)
g,n (p) = (−2pe dpe )
where eS0 is a proxy for N held fixed in the double-scaling |Aut(Γ)|
Γ∈Gg,n
∞ e∈EΓ
limit, corresponding to the inverse string coupling in (3).
Y  b(−1)mv
2gv −2+nv
The multi-resolvents Rg,n (x) are multi-valued func- × √
tions with branch cuts supported on the eigenvalue dis- v∈VΓ
2 sin(πmv b2 )
tribution of the matrix. They are thus naturally defined Y√
on a multi-sheeted cover of the complex plane with cuts × 2 sin(2πmv bpj )Vg(b)
v ,nv
(ipv ) . (13)
on the real axis, which defines a Riemann surface known j∈Iv

as the spectral curve. In the cases of interest, we can find

The sum runs over all possible degenerations of the world-
a rational parameterization (x(z), y(z)) ∈ C2 for z ∈ C.
sheet as nodal surfaces. They are combinatorially la-
The multi-resolvents (8) are determined recursively in
belled by stable graphs whose vertices v ∈ V indicate the
terms of the spectral curve via topological recursion [22–
different stable components of the surface of type (gv , nv )
24], which efficiently solves the loop equations of the ma-
while e ∈ E runs over the internal edges to which we as-
trix model. To formulate topological recursion one views
sign a momentum pe that we integrate out. We also as-
the resolvents as differentials on the spectral curve
sign a color mv ∈ Z≥1 to each vertex over whose choice we
(b)
ωg,n (z) = Rg,n (x(z)) dx(z1 ) · · · dx(zn ) , (9) also sum. The quantity Vg,n (ip) is a polynomial of degree
3g −3+n in p2j which remarkably is precisely the analytic
except for (g, n) = (0, 1), (0, 2) where the definition is continuation of the corresponding string amplitude of the
slightly modified. Virasoro minimal string, which constitutes a quantum
The spectral curve. Our central claim is that CLS is deformation of the Weil-Petersson volume [5]. Finally,
dual to a double-scaled two-matrix integral characterized the prime on the integral denotes integration from 0 to
by the following spectral curve ∞ in the complex plane, computed by decomposing the
√ √ trigonometric functions into exponentials and discarding
x(z) = −2 cos(πb−1 z), y(z) = 2 cos(πb z) , (10) divergent contributions without exponentials. We inter-
pret the stable graphs, corresponding to degenerations of
where z is a uniformizing coordinate. This spectral curve the worldsheet, as Feynman diagrams of the closed string
leads to richer resolvents than that of the (p, q) minimal field theory of CLS in a suitable gauge, with the quantum
string, since it (i) is not algebraic, and (ii) exhibits in- (b)
volumes Vg,n as the string vertices.
∗ ∗
finitely many branch points (zm such that dx(zm ) = 0) Checks. The expression (13) can be explicitly writ-
together with (iii) infinitely many nodal singularities ten out for low (g, n), but the number of stable graphs
(points z ± that are mapped to the same point on the grows rapidly for large (g, n). For (g, n) = (0, 3), one
spectral curve). can directly verify agreement with the worldsheet defini-
tion (2) which does not involve any moduli integration.
 4

For (g, n) = (0, 4) and (1, 1), the expression (13) agrees Examples of trumpet partition functions of particu-
(b)
with the result from the analytic bootstrap. For higher lar physical significance (a) reduce Zg,n (Ψ) to the string
(g, n), we proved that (13) has the correct poles (4) and amplitudes (2), (b) relate string amplitudes to matrix
discontinuities as computed from the worldsheet, obeys resolvent differentials (9), (c) compute thermal partition
the symmetries (5) and satisfies the dilaton equation (6). functions of the matrix model, and (d) contribute to the
The duality symmetry (5a) is very non-trivial and is re- non-perturbative effects mediated by ZZ-instantons:
lated to the x − y symmetry of topological recursion [25].
Recursion relation. The topological recursion of ωg,n (b) (b) δ(p − q)
(b)
Ztr (q, p) = A0,2 (q, p) = − (trivial) , (15a)
can be translated to the string amplitudes Ag,n via the 2p
relation (12). This leads to a recursion relation similar (b) √ dz
Ztr (z, p) = π sin(2π zp) √ (resolvent) , (15b)
to Mirzakhani’s recursion relation for the Weil-Petersson z
volumes [26], except for a non-trivial sphere three-point (b) 2b
(b) Ztr (β, p) = sin(2πbp)K2bp (2β) (thermal) , (15c)
amplitude A0,3 and a deformed integral kernel Hb , which π
also appeared previously in [5]. (b) 2 sin(2πrbp) sin(2πsb−1 p)
Ztr ((r, s), p) = (ZZ) . (15d)
p
WORLDSHEET BOUNDARIES AND Non-perturbative effects. The genus expansion (3) is
NON-PERTURBATIVE EFFECTS an asymptotic series that admits non-perturbative cor-
rections of order exp(#gs−1 ) = exp(# exp S0 ). These
The closed string amplitudes (2) are contrived observ- corrections can be systematically computed from open
ables from the perspective of the dual matrix model. worldsheets with ZZ-boundary conditions [28–30]. A
More natural observables are thermal partition functions complete basis of ZZ-instanton boundaries is spanned by
(which admit a holographic interpretation in terms of the ZZ(r,s) × ZZ(1,1) boundary condition [31] in the two
a 2d gravity path integral on surfaces with asymptotic Liouville factors with r, s ∈ Z≥1 . In particular, the ten-
boundaries) or resolvent differentials (9) associated with sion of the ZZ-instanton is computed from the empty
one of the matrices. We denote such more general ob- disk diagram, which can be extracted from the dilaton
(b)
servables by Zg,n (Ψ), where Ψ = (Ψ1 , . . . , Ψn ) is a set of equation (6) and the ZZ-trumpet (15d),
boundary conditions tied to the specific observable.
Gluing conformal boundaries. The computation of (b) sin(πrb2 ) sin(πsb−2 )
Tr,s = 8eS0 (−1)r+s . (16)
(b)
Zg,n (Ψ) can be encapsulated within a broader framework b−2 − b2
involving conformal boundary conditions for the string (b)
worldsheet CFT, labelled by a wavefunction Ψ(p+ , p− ) Notably, Tr,s is purely imaginary. Therefore, these non-
(b)
in the product of the two Liouville theories. The compu- perturbative effects of order exp(−Tr,s ) are highly os-
(b)
tation of Zg,n (Ψ) can then be reduced to the computation cillatory, instead of exponentially suppressed. This also
(b) implies via resurgence that the large order behaviour of
of Ag,n . This is in principle true for any critical world- (b)
sheet string theory but takes a remarkably simple form Ag,n as g → ∞ is alternating and the effective string
in the context of 2d or minimal string theories, where the coupling is imaginary, see also [32].
space of physical states is particularly simple. In the case The technology of gluing trumpets can be applied to
of CLS, the relation reads ZZ-instantons. An exception is the cylinder with iden-
tical instanton boundary conditions, where the integral
Z Y
n in (14) diverges due to the breakdown of Siegel gauge in
(b)
Z(b)
g,n (Ψ) = (−2pj dpj Ztr (Ψj , pj )) A(b)
g,n (p) . (14) the open string zero-mode sector. This divergence can
j=1 be resolved using string field theory [33–35].
On the two-matrix model side, non-perturbative cor-
Thus, all information in this broader class of observables
(b) rections are computed by integrating a single eigenvalue
Zg,n (Ψ) is entirely captured by the elementary observ- pair over a steepest descent contour of the effective po-
(b) (b)
able Ztr (Ψ, p) ≡ Z0,2 (Ψ, p). This corresponds to the tential, with saddle points corresponding to the world-
once-punctured disk diagram with Ψ boundary condition, sheet ZZ-instantons. The technology for saddle point
which is simply proportional to the on-shell wavefunction evaluation in matrix models is well-established [36, 37]
Ψ(p+ = p, p− = ip). In the context of 2d quantum grav- and reproduces the leading non-perturbative corrections
ity, this is commonly referred to as the ‘trumpet’ parti- (b)
to An (S0 ; p) (3) as computed diagrammatically from
tion function [6]. (14) means that a conformal boundary the string worldsheet in arbitrary multi-instanton sec-
can be interpreted as a particular smearing of bulk ver- tors. This analysis probes non-perturbative aspects of
tex operators of the string worldsheet theory. Different the duality.
worldsheet boundary conditions which lead to the same Non-perturbative completion. On the matrix model
trumpet partition function are ‘BRST-equivalent’ [27]. side, the locations of the instanton saddles coincide with
 5

the zeros of the density of states (11). This indicates with SL(2, C) Chern-Simons theory. Canonical quantiza-
that (11) can only be trusted up to its first zero at tion of the phase space on the initial value surface Σg,n
E0 = 2 cos(πb−2 ), after which the integration contour with a suitable gravitational polarization as in [46] yields
over the eigenvalue pairs is deformed into the complex (b)
a Hilbert space Hg,n (p) spanned by products of left-
plane. Different choices for this contour result in in- and right-moving conformal blocks of central charge c ∈
equivalent non-perturbative completions. This effectively 13+iR+ and conformal weights hj ∈ 21 +iR corresponding
yields a model whose density of states is supported on a to principal series representations of SL(2, C). The imag-
finite interval, reminiscent of [38–40]. inary part of the central charge is set by the dS radius in
3ℓdS
Planck units [47], Im c = 2G N
. Conformal weights are re-
lated to the masses of massive particles
p via the standard
DE SITTER QUANTUM GRAVITY
holographic dictionary, h = 21 ± 12 1 − m2 ℓ2dS [48]. In
gravity, we also gauge large diffeomorphisms, thus wave-
Sine dilaton gravity. A field redefinition recasts the (b)
functions Ψ ∈ Hg,n (p) must be crossing symmetric in the
worldsheet action of CLS (1) as the following dilaton sense of 2d CFT. Thus wavefunctions transform precisely
gravity action [5, 41–44] as 2d CFT correlators with central charge c ∈ 13 + iR
Z under diffeomorphisms and Weyl transformations.
i √

Sdil [Φ, g] = 2 d2 x g ΦR − π −1 sin(2πΦ) , (17) For de Sitter quantum gravity, we have to specify the
2b Σg,n (b)
wavefunction of the universe Ψ ∈ Hg,n (p) living at future
+
where Φ is the dilaton and we omit boundary contribu- infinity I . There is no clear principle telling us what
tions. The reciprocal of the prefactor can be interpreted Ψ should be. In particular, the analogue of the Hartle-
as the gravitational coupling constant, with a semiclas- Hawking state [49] neither seems to be well-defined since
sical limit as −ib2 → 0. The CLS string amplitudes are the sum over topologies diverges badly, nor does it seem
reinterpreted as a gravitational path integral to be normalizable [50, 51]. Instead, one can motivate
Z the Liouville correlator Ψ = hVp1 · · · Vpn i to be a valid
[Dg][Dg Φ] −Sdil [Φ,g] wavefunction of the universe. It can be uniquely boot-
A(b)
g,n (p) ≡ e Vp1 · · · Vpn , (18)
voldiff strapped by requiring normalizability of the wavefunction
and appropriate factorization under degeneration of I + .
with the vertex operators defined previously. Solutions String amplitudes as cosmological correlators. Given
to the equations of motion of (17) come in a family of (b)
Ψg,n ∈ Hg,n , we can compute cosmological correlators of
alternating dS and AdS vacua, non-dynamical massive particles by integrating over the
n
X metrics of I + [52],
R∗ = 2(−1)m − 4π (1 + 2bpj )δ 2 (ξ − ξj ) (19a) Z
j=1 kΨg,n k2 = |Ψg,n |2 . (20)
m Mg,n
Φ∗ = , m∈Z, (19b)
2
This recovers the amplitudes (2) up to normalization.
somewhat reminiscent of [45]. Here ξj are the locations (b)
de Sitter/matrix model correspondence. Ag,n com-
of the vertex operator insertions. The dS solutions ne- putes cosmological correlators of massive particles in a
cessitate a (−, −) signature of the metric, similar to the specific state of the universe. Given the correspondence
dS saddles of [32]. m can be identified with the color with the matrix model, this constitutes a precise version
mv appearing in the exact answer (13), which suggests of holography where the bulk observables are cosmologi-
that only the saddles m ≥ 1 contribute to the path in- cal correlators. Since we already integrate over all met-
tegral (18). (13) also suggests that we should consider rics in (20) and topology fluctuates in quantum gravity,
more general saddles consisting of degenerated Riemann we should actually consider the genus resummed ampli-
surfaces connected at nodes whose different components tudes (3) as the true observables of the bulk theory. The
v ∈ V can be in different vacua mv ∈ Z>0 . These rep- parameter S0 is a stand in for log N , where N measures
resent transition amplitudes between universes with dif- the degrees of freedom in the matrix model, and thus
ferent cosmological constants. Evaluating the path inte- plays the role of the de Sitter entropy.
gral (18) beyond the leading saddle we uncover the same
fluctuation theory as in the Virasoro minimal string [5].
Together this provides a path integral interpretation of ACKNOWLEDGEMENTS
the sum over stable graphs (13).
Liouville as the wavefunction of the universe. The We would like to thank dio Anninos, Aleksandr Arte-
string amplitudes can be reinterpreted as cosmological mev, Mattia Biancotto, Matthias R. Gaberdiel, Davide
correlators in dS3 quantum gravity. The Hilbert space Gaiotto, Alessandro Giacchetto, Victor Godet, Victor
of dS3 quantum gravity can be obtained via its relation Gorbenko, Clifford V. Johnson, Shota Komatsu, Adam
 6

Levine, Raghu Mahajan, Juan Maldacena, Alex Maloney, [14] J. Teschner, Phys. Lett. B 363, 65 (1995),
Marcos Mariño, Nico Valdes-Meller, Mehrdad Mirbabayi, arXiv:hep-th/9507109.
Miguel Montero, Nikita Nekrasov, Hirosi Ooguri, Boris [15] J. Teschner, Class. Quant. Grav. 18, R153 (2001),
arXiv:hep-th/0104158.
Post, Maximilian Schwick, Ashoke Sen, Douglas Stan-
[16] A. Zamolodchikov, Int. J. Mod. Phys. A 19S2, 510 (2004),
ford, Jörg Teschner, Cumrun Vafa, Herman Verlinde, Ed- arXiv:hep-th/0312279.
ward Witten and Xi Yin for discussions. SC, LE and BM [17] A. A. Belavin and A. B. Zamolodchikov,
thank l’Institut Pascal at Université Paris-Saclay, with Theor. Math. Phys. 147, 729 (2006).
the support of the program “Investissements d’avenir” [18] V. A. Kazakov, Phys. Lett. A 119, 140 (1986).
ANR-11-IDEX-0003-01, and SC and VAR thank the [19] E. Brezin, M. R. Douglas, V. Kazakov, and S. H.
Kavli Institute for Theoretical Physics (KITP), which is Shenker, Phys. Lett. B 237, 43 (1990).
supported in part by grant NSF PHY-2309135, for hospi- [20] M. R. Douglas, “The two-matrix model,” in
Random Surfaces and Quantum Gravity (Springer
tality during the course of this work. VAR is supported US, Boston, MA, 1991) pp. 77–83.
in part by the Simons Foundation Grant No. 488653, by [21] C. Crnkovic, P. H. Ginsparg, and G. W. Moore,
the Future Faculty in the Physical Sciences Fellowship at Phys. Lett. B 237, 196 (1990).
Princeton University, and a DeBenedictis Postdoctoral [22] B. Eynard and N. Orantin,
Fellowship and funds from UCSB. SC is supported by Commun. Num. Theor. Phys. 1, 347 (2007),
the U.S. Department of Energy, Office of Science, Of- arXiv:math-ph/0702045.
[23] B. Eynard, JHEP 01, 051 (2003), arXiv:hep-th/0210047.
fice of High Energy Physics of U.S. Department of En-
[24] L. Chekhov, B. Eynard, and N. Orantin,
ergy under grant Contract Number DE-SC0012567 (High JHEP 12, 053 (2006), arXiv:math-ph/0603003.
Energy Theory research), DOE Early Career Award DE- [25] B. Eynard and N. Orantin, (2007),
SC0021886 and the Packard Foundation Award in Quan- arXiv:0705.0958 [math-ph].
tum Black Holes and Quantum Computation. BM grate- [26] M. Mirzakhani, Invent. Math. 167, 179 (2006).
fully acknowledges funding provided by the Sivian Fund [27] N. Seiberg and D. Shih, JHEP 02, 021 (2004),
at the Institute for Advanced Study and the National arXiv:hep-th/0312170.
Science Foundation. [28] J. Polchinski, Phys. Rev. D 50, R6041 (1994),
arXiv:hep-th/9407031.
[29] B. Balthazar, V. A. Rodriguez, and X. Yin,
JHEP 05, 048 (2023), arXiv:1907.07688 [hep-th].
[30] B. Balthazar, V. A. Rodriguez, and X. Yin,
JHEP 05, 050 (2023), arXiv:1912.07170 [hep-th].
[1] D. J. Gross and A. A. Migdal, [31] A. B. Zamolodchikov and A. B. Zamolodchikov, , 280
Phys. Rev. Lett. 64, 127 (1990). (2001), arXiv:hep-th/0101152.
[2] M. R. Douglas and S. H. Shenker, [32] J. Cotler and K. Jensen, (2024),
Nucl. Phys. B 335, 635 (1990). arXiv:2401.01925 [hep-th].
[3] E. Brezin and V. A. Kazakov, [33] A. Sen, JHEP 08, 075 (2020),
Phys. Lett. B 236, 144 (1990). arXiv:2002.04043 [hep-th].
[4] N. Seiberg and D. Shih, [34] A. Sen, JHEP 11, 077 (2021),
Comptes Rendus Physique 6, 165 (2005), arXiv:2101.08566 [hep-th].
arXiv:hep-th/0409306. [35] D. S. Eniceicu, R. Mahajan, C. Murdia, and A. Sen,
[5] S. Collier, L. Eberhardt, B. Mühlmann, and JHEP 07, 139 (2022), arXiv:2202.03448 [hep-th].
V. A. Rodriguez, SciPost Phys. 16, 057 (2024), [36] J. M. Daul, V. A. Kazakov, and I. K. Kostov,
arXiv:2309.10846 [hep-th]. Nucl. Phys. B 409, 311 (1993), arXiv:hep-th/9303093.
[6] P. Saad, S. H. Shenker, and D. Stanford, (2019), [37] D. S. Eniceicu, R. Mahajan, C. Murdia, and A. Sen,
arXiv:1903.11115 [hep-th]. JHEP 10, 065 (2022), arXiv:2206.13531 [hep-th].
[7] S. Collier, L. Eberhardt, B. Mühlmann, and V. A. Ro- [38] M. Berkooz, M. Isachenkov, V. Narovlansky, and G. Tor-
driguez, “The complex Liouville string: the worldsheet,” rents, JHEP 03, 079 (2019), arXiv:1811.02584 [hep-th].
(to appear). [39] V. Narovlansky and H. Verlinde, (2023),
[8] S. Collier, L. Eberhardt, B. Mühlmann, and V. A. arXiv:2310.16994 [hep-th].
Rodriguez, “The complex Liouville string: the matrix [40] H. Verlinde and M. Zhang, (2024),
model,” (to appear). arXiv:2402.02584 [hep-th].
[9] S. Collier, L. Eberhardt, B. Mühlmann, and V. A. [41] T. G. Mertens and G. J. Turiaci, JHEP 01, 073 (2021),
Rodriguez, “The complex Liouville string: worldsheet arXiv:2006.07072 [hep-th].
boundaries and non-perturbative effects,” (to appear). [42] A. Goel, Investigations of Holographic Duality in Two
[10] S. Collier, L. Eberhardt, and B. Mühlmann, “The com- Dimensions, Ph.D. thesis, Princeton U. (2022).
plex Liouville string: de Sitter quantum gravity,” (to [43] A. Blommaert, T. G. Mertens, and S. Yao, (2023),
appear). arXiv:2312.00871 [hep-th].
[11] N. Seiberg, Prog. Theor. Phys. Suppl. 102, 319 (1990). [44] A. Blommaert, T. G. Mertens, and J. Papalini, (2024),
[12] H. Dorn and H. J. Otto, Nucl. Phys. B 429, 375 (1994), arXiv:2404.03535 [hep-th].
arXiv:hep-th/9403141. [45] D. Anninos and D. M. Hofman,
[13] A. B. Zamolodchikov and A. B. Zamolodchikov, Class. Quant. Grav. 35, 085003 (2018),
Nucl. Phys. B 477, 577 (1996), arXiv:hep-th/9506136. arXiv:1703.04622 [hep-th].
 7

[46] H. L. Verlinde, Nucl. Phys. B 337, 652 (1990). [49] J. B. Hartle and S. W. Hawking,
[47] J. D. Brown and M. Henneaux, Phys. Rev. D 28, 2960 (1983).
Commun. Math. Phys. 104, 207 (1986). [50] A. Castro and A. Maloney, JHEP 11, 096 (2012),
[48] A. Strominger, JHEP 10, 034 (2001), arXiv:1209.5757 [hep-th].
arXiv:hep-th/0106113. [51] V. Godet, (2024), arXiv:2405.09833 [hep-th].
[52] J. M. Maldacena, JHEP 05, 013 (2003),
arXiv:astro-ph/0210603.